— Software Engineer · London

Solomon
Wheeler

Software engineer at Nomura with a first-class degree in Computer Science from the University of Bath (83.7%). I build agentic AI systems, Python data pipelines, and full-stack TypeScript. My undergraduate dissertation was published as a research paper at KONVENS 2025.

Scroll to explore

— Experience

Career

solomon@portfolio:~/career
solomon@portfolio:~/career$ git log --oneline --graph --all
 
* 9f2a3e8 (HEAD -> main) Nomura · Software Engineer
| Jul 2024 – Present
| • Python ETL pipelines for financial data transformation
| • Agentic AI (ReAct / Plan & Execute) — LangChain / LangGraph / LangSmith
| • AWS Bedrock model provider; vector DB position comparison
| • TypeScript microservices & microfrontends; mentoring
 
* 7c8b1d4 Nomura · Industrial Placement
| Jun 2023 – Jul 2024
| • IPV system rewrite in TypeScript — first on Kubernetes in team
| • End-to-end features: NestJS, React, SQL, Jenkins / GitLab CI
| • Agile delivery; TDD and infrastructure as code
 
* 5b3c7d2 Infinity Tuition · Mathematics Tutor
| Jun 2021 – Dec 2023
| • GCSE–A-level tutoring; adapted to individual learning styles
 
* 2e9f5a1 (tag: education) University of Bath · BSc Computer Science
2021 – 2025
• First Class Honours · 83.7% · Top 20 in Year
• Dissertation published at KONVENS 2025: LLM Chain Routing
• DSA 98% · RL 97% · ML 95% · Discrete Maths 95%
 
solomon@portfolio:~/career$

— Listening

Currently playing

— Research

Publications

PublishedKONVENS 2025ACL Anthology

Reducing Environmental Costs whilst Maintaining Operational Effectiveness in Large Language Models through Chain Routing

Developed a routing system that intelligently directs LLM queries to the most cost- and energy-efficient model whilst preserving output quality — reducing computational waste without sacrificing capability. Implemented in Python using HuggingFace Transformers and PyTorch, and Dockerised for deployment on HPC infrastructure with workloads distributed across multiple GPUs. A version of my undergraduate dissertation.

PythonHuggingFacePyTorchDockerHPC
Read Paper ↗

— Work

Projects

Published · KONVENS 2025

LLM Chain Routing

My undergraduate dissertation, published as a research paper at KONVENS 2025. Developed a routing system that directs LLM queries to the most cost- and energy-efficient model whilst preserving output quality. Implemented in Python using HuggingFace Transformers and PyTorch, Dockerised for HPC deployment across multiple GPUs.

PythonPyTorchHuggingFaceDockerHPC
Read Paper ↗
Highest Mark in Year

Teaching Robots to Walk

Implemented five deep RL algorithms (PPO, DPO, DDPG, TD3, SAC) from scratch in Python to train a quadrupedal MuJoCo robot. Our agents learned to sprint using emergent two-legged locomotion, far exceeding the baseline walking objective.

PythonPyTorchMuJoCoReinforcement Learning

Graph & Minimum Spanning Tree

Visualisation tool for directed and undirected graphs — randomly generated or user-defined. Uses Prim's algorithm to find the minimum spanning tree from a specified node.

AlgorithmsGraph TheoryVisualisation
View on GitHub ↗

Mental Health Hackathon

Write messages to others when you're feeling good, receive them when you're having a rough day. Includes mood tracking over time to visualise progress. Built at a university hackathon.

HackathonWeb AppMental Health
View on Devpost ↗

Timebox Day Planner

A clean application for planning your day using the timeboxing technique. Built with Next.js, Tailwind CSS, and Shadcn UI.

Next.jsTailwind CSSShadcn UI
Live Demo ↗

— Skills

Tech Stack

Languages

Python, TypeScript & SQL primary

PythonTypeScriptSQLJavaC / C#

ML & AI

Production & research experience

PyTorchHuggingFace TransformersLangChain / LangGraphLangSmithAWS Bedrockscikit-learnVector Databases

Web & Backend

Full-stack TypeScript

ReactNext.jsNestJSREST APIsTailwind CSS

Infrastructure & DevOps

CI/CD, containers & observability

KubernetesDockerJenkinsGitLab CIGitHub ActionsPrometheusGrafanaVercel

— Education

Qualifications

University of Bath — BSc Computer Science with Placement

83.7% overall · 2021–2025

First Class HonoursTop 20 in Year

A-levels: Physics (A*), Mathematics (A*), Computer Science (A*) · The Emmbrook School

Key Modules

Data Structures & Algorithms98%
Reinforcement Learning97%
Discrete Mathematics & Databases95%
Machine Learning95%
Artificial Intelligence84%
Foundations of Computation84%
Human-Computer Interaction & UX84%
Extended Systems Project82%
Computer Systems Architecture 1 & 281%
Principles of Programming 1 & 281%
Mathematics for Computation76%
Visual Computing75%

— Open Source

GitHub

Fetching stats…

— Contact

Get in touch

Send me a message directly from here.